I run the ads. You keep the relationship.

An Italian grocery brand in a category locked up by established sellers with years of ranking history behind them. I took the account over in March 2025, after six months in which it had spent almost exactly as much as it had earned.

A blended 95% ACOS across the half-year before I inherited it, and 129% in the final month - a euro of spend for a euro of revenue, on grocery margins. The brief pulled in two directions as well: the client's stated priority was visibility first and sales second. That is a reasonable thing to want, and an expensive thing to want, in a category where the incumbents can outbid a challenger indefinitely.

The rebuild was about measurement, not tidiness. The account was fragmented until every distinct buying behaviour had its own place to sit and its own budget to defend - granular enough that it could prove things rather than argue about them. The large mayority of the keyword targeting was non-branded - in a category owned by incumbents, the brand's own name was never where the growth was going to come from.

That mattered specifically because of the visibility brief. The client wanted to push into formats I had doubts about. Rather than resist it, I built an account that could answer the question with evidence instead of opinion - so the push went ahead, properly funded and properly isolated.

It didn't work. It never worked. In the month it ran unchecked, spend more than doubled and ad sales fell. But because it was isolated, the failure was provable in four weeks rather than diffused across the account for a year - and once proven, it could be contained. The client kept asking for visibility. Each time it ran, it was killed earlier, and the conversion side was pushed harder to carry it.

Across the six months before I took over, the account ran at a blended 95% ACOS. Across the six months after, 38.8% - with ad sales up 6× against a 2.5× increase in spend. Six months either side, everything included, nothing selected.

The sharpest evidence sits in two adjacent months. In the one month the visibility formats weren't running, the account returned at 22.5% ACOS. The following month, with them running unchecked, it returned at 68.1% - on more than double the spend. The account proved its own client's brief wrong, in writing, in four weeks.

By August it was carrying more than three times its previous monthly spend at a third of its previous ACOS - while still absorbing the visibility spend the client insisted on. That is the part worth noticing. The search build was strong enough to win with a hand tied behind its back.

An Australian beauty brand, inherited in March 2025. The client wanted growth without giving up profitability - the thing every advertiser asks for and few accounts are built to deliver.

The account wasn't broken, and that was the awkward part. Nothing was on fire, the numbers looked defensible, and there was no obvious villain to point at. Accounts like this are where money disappears quietly - not through catastrophe, but through a long tail of spend nobody has had the time to interrogate. The temptation on inheriting one is to rebuild it. It looks decisive, it justifies the handover, and it is very often the wrong call.

I didn't restructure it. The account's bones were sound. What it lacked was someone willing to be ruthless about what wasn't working and patient about what was.

The first month was a subtraction. Unprofitable campaigns and targets came out, and the account spent 16% less as a result - while ad sales fell under 4%. That gap is the whole argument: almost everything I removed had been buying almost nothing. It also settled the question before any money went back in, which meant the growth that followed was built on a base that had been proven rather than assumed.

After that, the work was unglamorous and continuous - fund what converts, cut what doesn't, and do it weekly rather than quarterly. There was no clever restructure. There was attention.

The cleanest measure is a like-for-like month. January, before I took over, ran at 23.7% ACOS. September ran at 15.0% - on 14% less spend and 36% more ad sales.

Across the seven months I ran it, blended ACOS improved from 22.2% to 17.6%, with monthly ad sales averaging 27% higher on an unchanged average monthly spend.

More profitable, at greater volume, for the same money - and structurally the same account I inherited.

A US beauty brand launching several new products into one of the most aggressively contested categories on the platform, against established sellers with deep budgets and years of ranking history.

New listings: no reviews, no rank, no conversion data - and a brand almost nobody was searching for yet, which meant virtually everything had to be won on non-branded traffic, at non-branded prices. Margins were thin enough that a sustained triple-digit ACOS would have ended the launch before it started. The first month of real traffic came in at 101% ACOS - every dollar of revenue costing a dollar of ad spend.

The first month was deliberately underfunded. In a category this expensive, buying data at scale before you know what converts is the fastest way to burn a launch budget - so the account stayed small and bought its data cheaply while the listings were still unproven.

Budget went in only once the account had earned it - behind the products that had demonstrated they could convert, not spread evenly across the range. From there the work compounded: each month's increase in spend was funded by the efficiency the month before had produced, so the account was never scaling on borrowed confidence.

ACOS fell from 101% to 31% over six months. From the first full month at scale, ad sales grew 3.6× while spend grew 2.2× - the account was buying substantially more revenue per dollar at the top than at the bottom.

The final two months are worth being straight about. Efficiency stopped improving and settled around 31%, and the category softened. I kept pushing spend through both - 19% more across those two months - because the account was profitable at that level, and the alternative was protecting a percentage at the cost of volume the brand needed. That was a decision, not a drift.

A German kitchen brand competing against products that were cheaper and better reviewed, on a budget the client had decided - firmly, and permanently - not to increase. I'd already been running the account for over a year when this eight-month window opened.

Two structural disadvantages at once: priced above the competition, and reviewed below it. The budget was small enough that broad coverage was never on the table, and that wasn't negotiable. What the product did have was a genuine reason to exist - it was better than its reviews suggested - which made the entire problem one of reach: find the narrow set of shoppers who'd understand that, and pay for nobody else.

The obvious move on a small budget is to ask for a bigger one. That had been ruled out, so the account had to be built as though scarcity were permanent - and that changes what optimisation means. It stops being about capturing demand and starts being about refusing it. The account deliberately ran fewer campaigns, not more, and got narrower rather than broader - every impression served to a shopper who'd compare price and reviews and walk away was an impression taken from someone who wouldn't.

Part of what made that work was one competitor in particular: despite a better price and better reviews, one of their products consistently lost sales on its own keywords to this brand. But that was only one piece of it, not the whole story.

A second piece was language. A meaningful share of the relevant non-branded search terms were German-specific phrasing - the kind of terms someone managing the account without fluent German would never think to target, let alone find. That wasn't a small gap; it was where a chunk of the cheap, high-intent traffic had been sitting the whole time.

The third piece was going deep on a handful of non-branded keywords relevant to only a few products in the category, this one included. Most accounts spread bids evenly across the obvious broad terms. Here, the return was in pushing hard on the narrow ones almost nobody else was fully committed to.

Eight months at a blended 10.8% ACOS - better than a 9× return - on a budget that sat in the same narrow band for seven of them.

The trend underneath shows in the frozen stretch's bookends: on 17% more spend than February, August returned 128% more ad sales - at 9.5% ACOS against 18.4%. The softer months in between tracked seasonality, not the account.

Then, in the final month, the client relented and the budget roughly doubled. The account absorbed it without a wobble: the biggest sales month of the run, still at an 8× return in the very month the money doubled. Discipline built under scarcity doesn't wash out when the budget finally arrives; it scales.